1 / 80

Agenda

Bridging Multi-Core and Distributed Computing: all the way up to the Clouds. D. Caromel , et al. 1. Background: OASIS, ActiveEon 2. ProActive Overview : Programming, Scheduling, Resourcing 3 . Use Case: Genomics 4. Cloud Seeding . Agenda. Parallelism+Distribution

faith
Télécharger la présentation

Agenda

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Bridging Multi-Core and Distributed Computing: all the way up to the Clouds D. Caromel, et al. 1. Background: OASIS, ActiveEon 2. ProActive Overview : Programming, Scheduling, Resourcing 3. Use Case: Genomics 4. Cloud Seeding Agenda Parallelism+Distribution with Strong Model: Speed & Safety

  2. Key Objectives • Parallel Programming Model and Tools • Badly needed • for the masses • for new architectures: Multi-Cores & Clouds • As Effective as possible: • Efficient • However Programmer Productivity is first KSF • For both Multi-cores and Distributed • Actually the way around • Handling of ``Large-scale’’: up to 4 000 so far

  3. 1. Background 1. Background

  4. OASIS Team & INRIA • A joint team, Now about 35 persons • 2004: First ProActive User Group • 2009, April: ProActive 4.1, Distributed & Parallel: From Multi-cores to Enterprise GRIDs

  5. OASIS Team Composition (35) • PostDoc (1): • Regis Gascon (INRIA) • Engineers (10): • Elaine Isnard (AGOS) • Fabien Viale (ANR OMD2, Renault ) • Franca Perrina (AGOS) • Germain Sigety (INRIA) • Yu Feng (ETSI, FP6 EchoGrid) • Bastien Sauvan (ADT Galaxy) • Florin-Alexandru.Bratu (INRIA CPER) • Igor Smirnov (Microsoft) • Fabrice Fontenoy (AGOS) • Open position (Thales) • Trainee (2): • Etienne Vallette d’Osia (Master 2 ISI) • Laurent Vanni (Master 2 ISI) • Assistants (2): • Patricia Maleyran (INRIA) • Sandra Devauchelle (I3S) • Researchers (5): • D. Caromel (UNSA, Det. INRIA) • E. Madelaine (INRIA) • F. Baude (UNSA) • F. Huet (UNSA) • L. Henrio (CNRS) • PhDs (11): • Antonio Cansado (INRIA, Conicyt) • Brian Amedro (SCS-Agos) • Cristian Ruz (INRIA, Conicyt) • Elton Mathias (INRIA-Cordi) • Imen Filali (SCS-Agos / FP7 SOA4All) • Marcela Rivera (INRIA, Conicyt) • Muhammad Khan (STIC-Asia) • Paul Naoumenko (INRIA/Région PACA) • Viet Dung Doan (FP6 Bionets) • Virginie Contes (SOA4ALL) • Guilherme Pezzi (AGOS, CIFRE SCP) • + Visitors + Interns Located in Sophia Antipolis, between Nice and Cannes, Visitors and Students Welcome!

  6. Startup Company Born of INRIA • Co-developing, Support for ProActive Parallel Suite • Worldwide Customers: Fr, UK, Boston USA Some Partners:

  7. Multi-Cores

  8. Symetrical Multi-Core: 8-ways Niagara II • 8 cores • 4 Native threads per core • Linux see 32 cores!

  9. Multi-CoresA Few Key Points • Not Shared Memory (NUMA) • Moore’s Law rephrased: Nb. of Cores double every 18 to 24 months • Key expected Milestones: Cores per Chips (OTS) • 2010: 32 to 64 • 2012: 64 to 128 • 2014: 128 to 256 1 Million Cores Parallel Machines in 2012 100 M cores coming in 2020 • Multi-Cores are NUMA, and turning Heterogeneous (GPU) They are turning into SoC with NoC: NOT SMP!

  10. 2. Overview ProActive Parallel Suite

  11. 2. Programming Optimizing • Parallel Acceleration Toolkit in Java: • Java Parallel Programming • + Legacy-Code + Wrapping and Control • Taskflow Scheduling • Resource Manager • Multi-Core + Distributed Open Source Used in production by industry

  12. OW2: Object Web + Orient Ware

  13. OW2: Object Web + Orient Ware LIU Jiangning (CVIC SE), Prof. MA Dianfu (BEIHANG UNIVERSITY), Prof. WEI Jun (ISCAS), Prof. JIA Yan (NUDT), Prof. WANG Huaiming (NUDT), Mr. YUCHI Jan (MOST), Jean-Pierre Laisné (BULL), Prof. HUAI Jinpeng (BEIHANG UNIVERSITY), Julie Marguerite (INRIA), ZHOU Minghui (PEKING UNIVERSITY), Stephane Grumbach (French Embassy), Hongbo XU (GMRC), ZHOU Bin (NUDT), Than Ha Ngo (French Embassy).

  14. Product: ProActive Parallel Suite Java Parallel Toolkit Multi-Platform Job Scheduler Resource Manager Used in Production Today: 50 Cores  300 Cores early 2010 • Strong Differentiation: • Java Parallel Programming + Integration + • Portability: Linux, Windows, Mac + • Versatility: Desktops, Cluster, Grid, Clouds = Perfect Flexibility

  15. ProActive Parallel Suite • Three fully compatible modules: Programming Scheduling Resource Management

  16. ProActive Contributors

  17. ProActive Programming: Active Objects 17

  18. ProActive Programming

  19. ProActive : Active objects JVM A A WBN! A ag =newActive (“A”, […], VirtualNode) V v1 = ag.foo (param); V v2 = ag.bar (param); ... v1.bar(); //Wait-By-Necessity JVM ag v2 v1 V Wait-By-Necessity is a Dataflow Synchronization Java Object Active Object Req. Queue Future Object Proxy Thread Request 19

  20. Standard system at Runtime: No Sharing NoC: Network On Chip Proofs of Determinism 20

  21. (2) ASP: Asynchronous Sequential Processes • ASP  Confluence and Determinacy • Future updates can occur at any time • Execution characterized by the order of request senders • Determinacy of programs communicating over trees, … • A strong guide for implementation, • Fault-Tolerance and checkpointing, Model-Checking, …

  22. Key Point: Locality will more than ever be Fundamental • Let the programmer control it • No global shared memory At user choice PGAS: Partitioned Global Address Space

  23. TYPED ASYNCHRONOUS GROUPS 23

  24. Broadcast and Scatter ag JVM c3 c3 c3 c3 c3 c3 c3 c1 c1 c1 c1 c1 c1 c1 c2 c2 c2 c2 c2 c2 c2 JVM JVM s s JVM • Broadcast is the default behavior • Use a group as parameter, Scattered depends on rankings cg ag.bar(cg); // broadcast cg ProActive.setScatterGroup(cg); ag.bar(cg); // scatter cg 24

  25. Dynamic Dispatch Group c4 c4 c4 c6 c6 c6 c5 c5 c5 c7 c7 c7 c8 c8 c8 c0 c0 c0 c9 c9 c9 c3 c3 c3 c1 c1 c1 c2 c2 c2 JVM Slowest ag cg JVM Fastest JVM ag.bar(cg); JVM 25

  26. Abstractions for Parallelism The right Tool to do the Task right

  27. ProActive Parallel Suite • Workflows in Java • Master/Workers • SPMD • Components • …

  28. Components: GCM Standard 28

  29. GridCOMP Partners

  30. Objects to Distributed Components IoC: Inversion Of Control (set in XML) A Example of component instance V Truly Distributed Components Typed Group Java or Active Object JVM 30

  31. From 2004 to 2008: • 2004 Grid Plugtests: Winner: Univ CHILE Deployed 560 Workers all over the world on a very heterogeneous infrastructure (no VO) • 2008 Grid Plugtests: KAAPI, MOAIS Grenoble: 3609 Nodes ACT, China: Beihang University, Beijing, China: 4329 Nodes

  32. Lille: 500(198) Grid 5000 - ALADDIN Orsay 1000(684) Nancy: 500(334) Lyon 500(252) Grenoble 500(270) Rennes 522(522) Toulouse 500(116) Bordeaux 500(198) Sophia Antipolis 500(434)

  33. Chinese Collaborations on Grid PlugTests • Professor Chi • Prof. Baoping Yan • Hosted the IV Grid Plugtests Grid@works 2007 • CNIC: Computer and Network Information Center • SCC AS: Super Computing Center of AS • Prof. Ji Wang • In EchoGrid, Chinese Leader of OW2 • NUDT: National Univ. of Defense Technology • PDL: Laboratory of Parallel & Distributed Processing

  34. Denis Caromel Infrastructure tested in Plugtests andin GCM Deployment Standard • Protocols: • Rsh, ssh • Oarsh, Gsissh • Scheduler, and Grids: • GroupSSH, GroupRSH, GroupOARSH • ARC (NorduGrid), CGSP China Grid, EEGE gLITE, • Fura/InnerGrid (GridSystem Inc.) • GLOBUS, GridBus • IBM Load Leveler, LSF, Microsoft CCS (Windows HPC Server 2008) • Sun Grid Engine, OAR, PBS / Torque, PRUN • Clouds: • Amazon EC2

  35. GCM Official StandardizationGrid Component Model • Overall, the standardization is supported by industrials: • BT, FT-Orange, Nokia-Siemens, NEC, • Telefonica, Alcatel-Lucent, Huawei …

  36. Denis Caromel Infrastructure tested in Plugtests andin GCM Deployment Standard Interoperability: Cloud will start with existing IT infrastructure, Build Non Intrusive Cloud with ProActive • Protocols: • Rsh, ssh • Oarsh, Gsissh • Scheduler, and Grids: • GroupSSH, GroupRSH, GroupOARSH • ARC (NorduGrid), CGSP China Grid, EEGE gLITE, • Fura/InnerGrid (GridSystem Inc.) • GLOBUS, GridBus • IBM Load Leveler, LSF, Microsoft CCS (Windows HPC Server 2008) • Sun Grid Engine, OAR, PBS / Torque, PRUN • Clouds: • Amazon EC2

  37. IC2D: Optimizing 39

  38. IC2D

  39. IC2D

  40. ChartIt

  41. Pies for Analysis and Optimization

  42. Video 1: IC2D OptimizingMonitoring, Debugging, Optimizing

  43. Scheduling & Resourcing 45

  44. ProActiveScheduling 46

  45. ProActive Scheduling Big Picture • Multi-platform Graphical Client (RCP) • File-based or LDAP authentication ProActive Scheduler • Static Workflow Job Scheduling, Native and Java tasks, Retry on Error, Priority Policy, Configuration Scripts,… ProActive Resource Manager • Dynamic and Static node sources, Resource Selection by script, Monitoring and Control GUI,… RESOURCES • ProActive Deployment capabilities: Desktops, Clusters, Clouds,…

  46. Scheduler: User Interface

  47. Another Example : Picture Denoising Split Denoise Denoise Denoise Denoise Merge • with selection on native executable availability (ImageMagik, GREYstoration) • Multi-platform selection and command generation • with file transfer in pre/post scripts Job

  48. ProActive Resourcing 50

More Related